A key component in many digital imaging systems is a digital printer. A digital printer produces hard copy output from a digital representation of an image. Digital printers have been made using many different basic technologies such as thermal dye-diffusion, electrophotography, ink-jet, and digital silver-halide writers. Such printers can either be monochrome (usually "black-and-white"), or may print multiple colors (typically cyan, magenta, and yellow). The input signal to a digital printer is a digitally encoded representation of the desired image. Typically, this includes a multi-level representation of the desired image density for each color plane at each x-y location (pixel) in the image. Alternatively, the information (such as a PostScript file) necessary to create such a representation can be supplied as input to the printer. Typically, each pixel may be characterized by an 8-bit digital value for each color plane in the image. These digital values are commonly referred to as code values. This provides 2.sup.8 =256 possible density levels for the digital printer to reproduce.
A digital printer will typically respond to some fundamental control parameter. This control parameter will vary depending on the particular output technology. For example, thermal dye-diffusion printers typically respond to the number of heat pulses applied by a heater element for a given pixel, and silver halide printers typically respond to the digitally controlled intensity of a laser spot. The value of the control parameter will be referred to as the printer control signal. It is possible to measure the image density formed on a digital printer as a function of the value of the printer control signal. For example the optical density of the image as a function of the number of thermal pulses can be measured for a thermal dye-diffusion printer. This relationship between the physical output response of the digital printer and the printer control signal will be referred to as the "raw sensitometry" of the digital printer.
Usually the raw sensitometry for a digital printer does not correspond to the desired output density as a function of the input signal to the digital printer. As a result it is frequently necessary to apply a "printer calibration function" to convert the input signal to the appropriate printer control signal. For example, the printer calibration function may convert an 8-bit input signal into the number of thermal pulses necessary to produce the desired output density for each value of the input signal. This is illustrated in FIG. 1 which shows a digital printer responding to an input signal I(x,y) for each x, y pixel of the image. A calibration function 12 is applied to the input signal I(x,y) to produce a printer control signal P(x,y). Often the calibration function may be incorporated into the digital printer itself so that it is not apparent as a separate component to the user as is shown in FIG. 2. In this case, the digital printer 20 includes a digital print engine 22 as well as a printer calibration function 24. For a digital color printer, there will usually be three or four input color channels. Each color channel will typically have it's own calibration function. FIG. 3 shows a three color printer 30 having red, green, and blue input signals given by I.sub.R (x,y), I.sub.G (x,y), and I.sub.B (x,y), respectively. The calibration function for a red color channel 34 processes the red input signal I.sub.R (x,y) to form a red printer control signal P.sub.R (x,y). Likewise, the calibration functions for the other two channels (36 and 38) are used to process the corresponding input signals. The printer control signals are then used to drive a digital print engine 32.
Fundamental to being able to determine a printer calibration function is the accurate knowledge of the digital printer's raw sensitometry. As a result, printer manufacturers will typically go to great lengths to characterize and control the raw sensitometry of the digital printer as closely as possible in the manufacturing process. In many cases, however, the raw sensitometry of a printer will vary over time due to factors such as media variability, aging of the digital printer's components, and changes in the digital printer's environment. If the printer calibration function is not modified accordingly, the output density formed by the printer will also vary over time. This can manifest itself as a change in the overall density of the image, or in the case of a color printer, as a change in the color-balance of the image. This last effect can be particularly objectionable due to the fact that color balance errors are more easily perceived by a human observer than density errors. As a result it is frequently desirable to be able to measure the raw sensitometry of a printer in the field so that an updated calibration table can be produced and used in the printer.
A number of prior art methods of determining the raw sensitometry for a digital printer involve printing a calibration target having patches created using a series of different printer control signal values (see U.S. Pat. No. 5,053,866). The raw sensitometry can then be determined by measuring the output density (or some other output quantity) using a densitometer (or some other measurement instrument). Typically it is not necessary or desirable to measure the output density for every possible value of the printer control signal. More often, some subset of the printer control signals are used, and the raw sensitometry values for the remaining printer control signals can be estimated using interpolation and smoothing methods. One problem with this method however is that the measured raw sensitometry function is quite susceptible to errors introduced by measurement noise, density variability (both within a print, as well as print-to-print), and image artifacts. As a result, the resulting printer calibration table produced from the raw sensitometry will contain errors as well. These errors can be particularly objectionable for color printers because of the fact that errors in determining the raw sensitometry in one color plane can result in color balance errors which vary across the tone scale. For example, if you were to print a smooth neutral gradient spanning the range from black to white, some portions of the gradient might appear to have a greenish cast, while others might appear to have a reddish cast. This will be quite objectionable to a human observer.
Typically errors in the raw sensitometry measurements can be minimized by performing many replications of the measurements, and subsequently applying statistical techniques to eliminate bad data points, and average out the measurement errors. Although this is useful in the determination of the calibration function during the printer manufacturing process, it is frequently not convenient to do this when updating the printer calibration in the field where it is desired to make the fewest number of prints and measurements, and to complete the calibration procedure in the shortest possible time.
Additionally, access to a densitometer or some other measuring device is required when using instrumented calibration techniques. Many times these devices may not be available when performing calibration of the printer in the field. A number of techniques have recently been disclosed (see U.S. Pat. Nos. 5,298,993 and 5,347,369) which teach the use of calibration targets that can be "measured" using only a human observer. These techniques, which will be referred to as "visual calibration techniques," also tend to be susceptible to noise in the visual judgment process. In fact, since the measurement variability tends to be larger in many cases, the errors can actually be substantially larger than those associated with instrumented measurements.